Randomised trial finds AI maternal health chatbot does not ease anxiety or depression

A randomised clinical trial published in JAMA Network Open on 2 November 2026 found that giving pregnant and postpartum people access to a generative AI maternal health chatbot did not reduce anxiety or depression at 12 months compared with an active control¹.

What the trial found

The trial enrolled 400 participants between October 2023 and December 2024 and completed 12-month follow-up in December 2025¹. Recruitment was online, across 49 US states and the District of Columbia. To be eligible, participants had to speak English, be aged 14 or older, be pregnant or parenting an infant under 6 months, and self-identify as a member of a racial or ethnic minoritised group¹.

Participants were allocated 1:1 to the chatbot or to a control group that received children's books each month¹. Mean age was 32.6 years, and 200 (50%) were already parenting an infant when they enrolled. Of the 400, 188 (47%) identified as Black or African American, 80 (20%) as Asian, 65 (16%) as Hispanic or Latino, 57 (14%) as multiracial and 10 (2%) as American Indian or Alaska Native¹.

The primary outcomes were anxiety and depression symptoms at 12 months, measured on the GAD-7 and PHQ-9 scales. The chatbot made no measurable difference to either. The adjusted mean difference was 0.10 for anxiety (95% CI, -0.77 to 0.96) and -0.23 for depression (95% CI, -1.05 to 0.59)¹.

Secondary outcomes were also flat. The chatbot showed no association with emergency department visits, postpartum check-ups, use of effective contraception or sleep duration¹. Among participants who were pregnant at enrolment, the chatbot group reported more pregnancy-related complications at the 12-month secondary end point, though this did not hold in the longitudinal analyses¹. In exploratory analyses within the chatbot group, heavier use was linked both to longer sleep and to higher anxiety scores¹.

What it means for the market

The authors note that generative AI chatbots are already in growing use for maternal health support while randomised evidence remains limited¹. This trial adds a null result on the outcomes that matter most to commissioners and payers: mental health, emergency attendance and uptake of postnatal care.

For health systems weighing patient-facing chatbots, the study separates use from benefit: within the chatbot group, more engagement did not line up with better mental health, and it was associated with higher anxiety scores in exploratory analysis¹. That gives buyers and vendors a randomised reference point when claims are made that a chatbot improves maternal outcomes rather than simply answering questions.

The study does not say which chatbot was used beyond its description as a generative AI maternal health tool, so the result cannot yet be read across to specific products¹.

Caveats

The trial was run in the US, in English speakers from minoritised groups recruited online, so it may not generalise to other populations or care settings¹. The control group received an active intervention rather than nothing. The authors themselves say the secondary and exploratory findings, including the signal on pregnancy complications and the link between engagement and anxiety, should be interpreted cautiously¹. The trial is registered as NCT06053515¹.